Effect of temperature experienced during development on wing traits. For each trait, we plot the mean value as a function of temperature and use bars to represent the standard deviation for 24 to 38 measurements per temperature. These representations, called reaction norms, are the standard way of displaying plasticity. Trait icons, compare with Figure 2, are given on the right of the respective reaction norm line: (A-B) dorsal eyespots, (C-F) ventral eyespots on forewing and hindwing, (G) ventral bands and (H) wing areas. We tested for the effect of temperature on wing pattern trait using the model trait ~ temperature + wing (where the area of the corresponding wing is a covariate) and on wing area using wing ~ temperature (see Material and methods). Trait values were used untransformed or log10 transformed to meet the Shapiro-Wilk normality test (alpha = 0.05). Statistical significance for effects of temperature on wing traits (see Material and methods) is indicated to the left of each reaction norm: ns (non-significant), P >0.01, **P <0.01, ***P <0.001. When ANCOVA/ANOVA showed significant effects of temperature on trait value, we compared across temperatures. For each reaction norm, different letters indicate pairwise comparisons that revealed statistically significant differences (lsmeans P <0.01) (see Additional file 1 for more details on these statistical analyses). For the number of white pupils (n = 30 to 38 individuals, Additional file 1) on the dorsal surface of the hindwing in panel (I), we found no significant effect of temperature using the model pupil nr ~ temperature with a quasi-Poisson distribution (Deviance = 1.894, df = 2, P = 0.1172). ANCOVA, analysis of covariance; ANOVA, analysis of variance.